UK Medicines Regulator Calls for New AI Laws in NHS

The UK's Medicines and Healthcare products Regulatory Agency (MHRA) has concluded that current law is failing to keep pace with the rollout of artificial intelligence across the National Health Service (NHS). An independent commission set up by the agency has published 44 recommendations for legislative changes covering AI tools already in use in British clinics and hospitals today.

What the commission proposes

The report, commissioned by the MHRA from an independent panel of experts, proposes among other things a system of staged certification for AI tools, replacing a single one-time approval that covers a product's entire life cycle. Under the new system, AI tools would earn successive levels of trust in stages, while the regulator would gain the power to continuously monitor already-deployed algorithms and revoke approval if their performance declines or they begin to malfunction.

The second pillar of the proposal is a system working-titled "L plate," a reference to the "L" plates fitted to cars driven by learner drivers in the UK. It would let medical staff test new AI models under close supervision before a tool is rolled out for full clinical use.

Penalties for faulty tools

The commission also proposes granting the regulator clear powers to penalize companies that develop AI-based medical software if their products fail to meet required safety and performance standards. This marks a shift from the current model, in which responsibility for faulty tools is diffused across the manufacturer, the healthcare provider, and the regulator.

The recommendations also include a patient's right to know whether artificial intelligence is involved in their care, along with access to information about the tools being used. A study by the University of Edinburgh, cited in the discussion around the new rules, found that patients who are aware AI is processing their data are less likely to share sensitive information with their doctor.

Why the old rules fall short

MHRA chief executive Lawrence Tallon stressed that the current legal framework for medical devices was built in an era of physical products, such as hip implants or stethoscopes, and was never designed for software that learns, adapts, and changes its behavior after it has already received market approval.

Patients will increasingly come to see AI as part of normal care within the NHS - Lawrence Tallon, MHRA chief executive

The commission, launched by the UK government in September 2025, is tasked with advising the MHRA on implementing commitments from the country's ten-year health plan. Its members include clinicians, regulators, and experts who have worked with companies including Google and Apple, and the report openly acknowledges that no country in the world has yet found a complete solution to regulating AI in medicine.

What it means for Poland and Europe

The UK is charting a different course here than the European Union, where the framework for AI in medicine is set by the EU AI Act, which will apply to healthcare providers from August 2, 2026. Still, the MHRA's proposals point in a similar direction: phased approval of AI tools for clinical use, ongoing oversight of their performance after deployment, and a clear patient right to know when an algorithm is involved in their care.

For Polish hospitals, where AI is already being used in imaging diagnostics and in describing X-ray or MRI scans, the British model could become a reference point in discussions about how to oversee tools that keep changing their behavior after deployment, not just at the point of certification. The scale of AI scribe use among British family doctors, reaching 40 percent, also shows how quickly practice can outpace regulation in countries that started adopting these tools earlier.